The calorimeter was moved downstream for the 2019 and 2021 runs, but the value of ECAL_FACE in tracking/src/main/java/org/hps/recon/tracking/BeamlineConstants.java was not updated. This means that when comparing the track state as returned by TrackStateUtils.getTrackStateAtECal(track) to the cluster position as returned by cluster.getPosition() they are at very different z locations:

This had been fixed in git branch iss914, but was not merged into the master branch before that issue was closed.